How to get the pixel values of specific object in an image?

Hi everyone! I would like to obtain the pixel values of a very thin and little streak (the one circled in red in this figure) . I have some problems because this object is very far. I have previously used this working code on a much larger and closer stripe, but for this strip, it doesn't seem to work. Can someone help me?

clc; clear all; close all;
A=rgb2gray(imread('Img_0222.jpg'));
B=bwareafilt(imbinarize(A-medfilt2(A,[5,5])),1);
pixelValues=A(B)
figure
imshow(B)

Answer 1

[Image Analyst]

Try this:

% Demo by Image Analyst
clc;    % Clear the command window.
close all;  % Close all figures (except those of imtool.)
clear;  % Erase all existing variables. Or clearvars if you want.
workspace;  % Make sure the workspace panel is showing.
format long g;
format compact;
fontSize = 20;
markerSize = 40;
%--------------------------------------------------------------------------------------------------------
%    READ IN IMAGE
fileName = 'white spots.jpg';
grayImage = imread(fileName);
% Get the dimensions of the image.
% numberOfColorChannels should be = 1 for a gray scale image, and 3 for an RGB color image.
[rows, columns, numberOfColorChannels] = size(grayImage)
if numberOfColorChannels > 1
    % It's not really gray scale like we expected - it's color.
    % Extract the blue channel (so the magenta lines will be white).
    grayImage = grayImage(:, :, 3);
end
%--------------------------------------------------------------------------------------------------------
% Display the image.
subplot(2, 2, 1);
imshow(grayImage, []);
impixelinfo;
axis('on', 'image');
title('Original Gray Scale Image', 'FontSize', fontSize, 'Interpreter', 'None');
hold on
drawnow;
% Maximize window.
g = gcf;
g.WindowState = 'maximized'
drawnow;
n = 47;
kernel = ones(n)/n^2;
blurredImage = imfilter(grayImage, kernel);
subplot(2, 2, 2);
imshow(blurredImage, []);
% Get a profile
horizontalProfile = improfile(blurredImage, [1, columns], [rows/2, rows/2]);
verticalProfile = improfile(blurredImage, [columns/2, columns/2], [1, rows]);
horizontalProfile = movmean(horizontalProfile, 199);
verticalProfile = movmean(verticalProfile, 199);
subplot(2, 2, 3);
cla
plot(horizontalProfile, 'b-', 'LineWidth', 2)
hold on;
plot(verticalProfile, 'r-', 'LineWidth', 2)
grid on;
backgroundImage = zeros(rows, columns);
for col = 1 : columns
    for row = 1 : rows
        backgroundImage(row, col) = sqrt(horizontalProfile(col)^2 + verticalProfile(row)^2);
    end
end
backgroundImage = rescale(backgroundImage, 0, 1);
subplot(2, 2, 2);
imshow(backgroundImage, []);
impixelinfo;
title('Background Image', 'FontSize', fontSize, 'Interpreter', 'None');
grayImage = uint8(double(grayImage) ./ backgroundImage);
%--------------------------------------------------------------------------------------------------------
subplot(2, 2, 4);
imshow(grayImage, []);
title('Background Corrected Image', 'FontSize', fontSize, 'Interpreter', 'None');
impixelinfo;
% grayImage = grayImage(1800:2200, 1800:2200);
figure;
mask = grayImage == 255;
mask = bwareafilt(mask, [60, inf]);
imshow(mask)
% Get long skinny blobs only.
props = regionprops(mask, 'Area', 'MajorAxisLength', 'MinorAxisLength');
% allAreas = sort([props.Area], 'descend')
aspectRatios = [props.MajorAxisLength] ./ [props.MinorAxisLength];
keepers = find(aspectRatios > 5);
labeledImage = bwlabel(mask);
mask = ismember(labeledImage, keepers);
% Re-analyze with new mask.
props = regionprops(mask, 'Area', 'MajorAxisLength', 'MinorAxisLength');
imshow(mask)
allAreas = sort([props.Area], 'descend')
aspectRatios = [props.MajorAxisLength] ./ [props.MinorAxisLength]
title('Final Binary Image', 'FontSize', fontSize, 'Interpreter', 'None');

[Gargolla9]

@Image Analyst thank you! Your code works also for other images that I have. However can you explain me in simple words what are we doing in the code?

[Image Analyst]

I read in the image and noticed that there is a lens shading -- an intensity falloff as you get to the sides and corners of the image. So that makes ir difficult to get a global threshold so I wanted to background correct the image to "flatten" it so we could get a global threshold. So I got the vertical and horizontal intensity profile through the middle of the image and then made an estimate of what the background would be. The background being the percentage of the light that hits each part of the image. Then I divide the image by the background to flatten it. If the light hitting the corner was only 90% of that hitting the middle, what should you do? Divide the values there by 90% of course.

Then I noticed that all the white spots and streaks seemed to be clipped at a value of 255 so I created a binary image by taking only those pixels with a value of 255. Then I threw out blobs smaller than 60 pixels in area. That left only a few spots and streaks. Then I found the major and minor axis lengths of each blob and divided them to get the aspect ratio. Spots will have an aspect ratio of around 1 or 2, and oblong objects will have an aspect ratio of around 5 or more. So I found out what blob numbers had high aspect ratios. I labeled the image to give each blob an "ID" number and I passed that list along with the list of blobs we want to keep into ismember(). That function returns only the oblong blobs we want. Then I just remeasured some parameters of only those desired oblong blobs.

[Gargolla9]

@Image Analyst wonderful job! just an info. Once you have extrapolated the streak in the final binary Image, how can I obtain the values of the pixels belonging only to that streak? Because it is the most important info that I need after extrapolating the streak

[Image Analyst]

From examining the gray scale image highly magnified, it looks like all the values are 255. Were you expecting something less? Looks like your image is overwxposed and saturated.

[Gargolla9]

I think that 255 Is acceltable, but I have to apply your code to other 30 frames of the same object (these frames are very similar to the One that I have posted). So I would like to understand in which part of the code you can extract the exact information about pixel values (At the end I have to plot pixel values of the streak in each frame vs UTC Gregorian)

[Image Analyst]

Once you have the final mask you can get the pixel values like this:

pixelvalues = grayImage(mask);

Those will be the background corrected image values. If you want the original values (darkening as it goes to the sides) then you'll have to save the original image or not overwrite it.

originalImage = grayImage; % Save a copy of the original image, before background correction.
grayImage = uint8(double(grayImage) ./ backgroundImage);
% Then more code to get the mask, then:
% Get original pixel values:
pixelvalues = originalImage(mask);
% Get background corrected pixel values.
pixelvalues = grayImage(mask);

Answer 2

[Image Analyst]

I'm not seeing a "stripe" in the middle of the image, just white spots. Can you outline it in red? Maybe you can sum the image horizontally and look at the profile.

verticalProfile = sum(A, 2);
plot(verticalProfile, 'b-', 'LineWidth', 2);
grid on;

[Gargolla9]

@Image Analyst Hi, I have circled the strike in red. Instead this here is the original Image. Please can you help me?

[Image Analyst]

Can you just all imbinarize, then imclearborder to get rid of the streaks that go all the way across the image (or get rid of them some other way) perhaps with the adaptive option, then call regionprops asking for the max and min axis lengths? Then find the ratio of max length to min length and keep only those with a ratio of about 2 or more?

[Gargolla9]

@Image Analyst Can you help me more practically with the code? Please, because I am new to matlab Image Processing

[Image Analyst]

I'm watching football now. No time for several hours. Really, it's not hard, try it yourself after you've gone through my Image Segmentation Tutorial here:

Image Analyst's File Exchange

[Image Analyst]

OK, so I'm sure by now you've called

mask = imbinarize(grayImage, 'adaptive');
props = regionprops(mask, 'MajorAxisLength', 'MinorAxisLength');
aspectRatios = [props.MajorAxisLength] ./ [props.MinorAxisLength];
keepers = aspectRatios > 2;
labeledImage = bwlabel(mask);
mask = ismember(labeledImage, keepers)
props = regionprops(mask, 'MajorAxisLength', 'MinorAxisLength');

So, how did it go?

Answer 3

[yanqi liu]

clc; clear all; close all;
A=rgb2gray(imread('https://www.mathworks.com/matlabcentral/answers/uploaded_files/791924/image.jpeg'));
A2 = imresize(A,0.2,'bilinear');
B2 = imbinarize(A2,'adaptive','ForegroundPolarity','dark','Sensitivity',0.7);
B3 = imopen(B2, strel('line', 7, 135));
[r,c] = find(B3);
B2 = bwselect(B2, c, r);
[L,num] = bwlabel(B2);
stats = regionprops(L);
for i = 1 : num
    recti = stats(i).BoundingBox;
    areai = stats(i).Area;
    if recti(3)/size(B2,2) > 0.05 || recti(4)/size(B2,1) > 0.05 || recti(4)/size(B2,1) < 0.01 || areai > 50
        B2(L==i) = 0;
        continue;
    end
    if recti(1)/size(B2,2) > 0.2 && recti(2)/size(B2,1) > 0.2 && (recti(1)+recti(3))/size(B2,2) < 0.8 && (recti(2)+recti(4))/size(B2,1) < 0.8
    else
        B2(L==i) = 0;
        continue;
    end
end
B2 = imdilate(bwperim(imdilate(B2, strel('disk', 9))), strel('disk', 2));
A31 = A2; A32 = A2; A33 = A2;
A31(B2) = 255;A32(B2) = 0;A33(B2) = 0;
A3 = cat(3,A31,A32, A33);
figure; imshow(A3);